Skip to content

Latest commit

 

History

History
73 lines (58 loc) · 1.67 KB

progress-reporter.md

File metadata and controls

73 lines (58 loc) · 1.67 KB

进度报告机制(ProgressReporter)

长时间训练任务可通过 ProgressReporter 向前端报告实时进度,前端通过轮询 /workspace/progress.json 获取进度信息并显示进度条。

模块位置:local-server/src/dmla_progress.py

使用方式:

from dmla_progress import ProgressReporter

# 初始化进度报告器
progress = ProgressReporter(
    total_steps=100,        # 总步数(如 epoch 数)
    description="训练 AlexNet"
)

# 训练循环中更新进度
for epoch in range(100):
    train_one_epoch()
    loss = calculate_loss()
    progress.update(
        step=epoch + 1,
        message=f"Epoch {epoch+1}: Loss={loss:.4f}",
        extra_data={"loss": loss, "accuracy": accuracy}
    )

# 训练完成
progress.complete(
    message="训练完成",
    extra_data={"final_accuracy": 0.85}
)

进度数据结构:

{
  "description": "训练 AlexNet",
  "total_steps": 100,
  "current_step": 50,
  "percent": 50.0,
  "message": "Epoch 50: Loss=0.32",
  "status": "running",
  "start_time": "2026-05-02T12:00:00Z",
  "elapsed_seconds": 120,
  "estimated_remaining": 120,
  "extra_data": {"loss": 0.32, "accuracy": 0.78}
}

状态类型:

  • starting: 任务开始
  • running: 任务进行中
  • complete: 任务完成
  • error: 任务出错

辅助函数:

from dmla_progress import get_progress, clear_progress

# 读取当前进度
progress = get_progress()

# 清除进度文件
clear_progress()

注意事项:

  • 进度文件路径固定为 /workspace/progress.json
  • 写入失败不会影响训练,仅打印警告
  • extra_data 可携带自定义指标(loss、accuracy 等)